Update README.md

README.md

@@ -181,7 +181,9 @@ python run_training_pipeline.py <shorthand of the pipeline>
 ```
 
 You can supply any of the following arguments, but don't have to (although for training you should definitely specify at
-least a GPU ID).
+least a GPU ID). With this version of the toolkit, we recommend that you download the meta-checkpoint of the model you want 
+to train and fine-tune from it to significantly decrease need for data, time for training and chance of success if your data 
+is not very high quality.
 
 ```
 --gpu_id <ID of the GPU you wish to use, as displayed with nvidia-smi, default is cpu> 
@@ -279,15 +281,6 @@ Here are a few points that were brought up by users:
 
 ---
 
-## Example Pipelines available
-
-| Dataset                                                                              | Language | Single or Multi | TransformerTTS | Tacotron 2 | FastSpeech 2 | 
-| -------------------------------------------------------------------------------------|----------|-----------------|:--------------:|:---------:|:-----------:|
-| [LJSpeech](https://keithito.com/LJ-Speech-Dataset/)                                  | English  | Single Speaker | ✅              | ✅        |✅           |
-| [Nancy Krebs](https://www.cstr.ed.ac.uk/projects/blizzard/2011/lessac_blizzard2011/) | English  | Single Speaker | ✅              | ✅        |✅           |
-
----
-
 This toolkit has been written by Florian Lux (except for the pytorch modules taken
 from [ESPnet](https://github.com/espnet/espnet) and
 [ParallelWaveGAN](https://github.com/kan-bayashi/ParallelWaveGAN), as mentioned above), so if you come across problems